Here is the third. For background, the mechanics of branching stories - the way that outcomes are determined - are based on user decisions. So it is natural for players to become overly intellectual in their selection of what to do or say next. This can be especially awkward if, after some dialogue and other context setting, the simulation suddenly presents a big decision.
What I do is train (i.e. desensitize) the player to the type of decision they will eventually be making. In one sim, they player had to ultimately decide if they were going to give the onscreen character some sensitive information or not. To set the player up, I gave them a bunch of short, no-consequence choices they had to make with the on-screen character leading up to the big one. (This is a different technique than using small decisions to gently impact underlying variables that later cumulatively impact available options and results.)
Here is an example of an early no-consequence decision that comes before a very consequential ask for information a minute or so later.
The interaction should be somewhat interesting, defining that character, moving the story along, feeling as if there might be consequences or other weight, or at the very least a venue for self-expression.
From an authoring perspective, the response of the on-screen character may take up one node before going back to the central path, or it may be a generic response on the central path.Also, of course, this interaction has to happen for every play, not as the result of a specific player choice.